We have used a novel approach of combining gene expression profiling and array CGH to develop a predictor of which gastric cancer patients are likely to respond to cisplatin and 5-FU chemotherapy (CF). This predictor is based upon the use of 96 tumors biopsied from patients with metastatic gastric cancer prior to the start of CF therapy for which detailed clinical follow-up was also obtained. A gene signature containing hundreds of genes distinguished CF responders from non-responders. In order to reduce this number of genes to a more clinically useful set of genes, we identified which genes were also contained in regions of genome amplification. Eventually, a 3-gene predictor was developed using the 96 tumor training set and validated using a separate validation cohort of 27 samples. PCR and immunohistochemistry further confirmed the over-expression of the three genes in the poor responders to CF therapy, suggesting that a clinical test could be developed to predict which patient may or may not respond to CF therapy. This work has been submitted for a patent and we have been approached by a biotech company to negotiate a CRADA for this technology. The predictor did perform as a prognostic predictor in a set of array data from gastric cancer patients who were treated only by surgical resection, nor treated by very different drug therapies. This strongly suggests that the predictor reflects the patient response to CF therapy and is not a general predictor of patient prognosis. This study is also extremely unique since we obtained second biopsies from many of the same patients who initially demonstrated a good response to CF therapy but eventually exhibited resistance to the therapy with continued progression of the tumors. The second biopsy was obtained following the documentation of tumor progression. The results of our analyses comparing the gene signature associated with initial tumor resistance to CF with the signature of acquired resistance to CF demonstrated significant overlap, suggesting that acquired resistance was the result of a selective outgrowth of tumor cells already in the initial tumor that were resistant to CF. We have also used proteomic approaches to identify about 20 proteins that are specific to gastric cancer tumors compared to normal stomach epithelial. These exciting results are being followed-up to determine whether any of these proteins could be useful as biomarkers to detect early gastric cancer or to follow disease progression in patients. The results of this study were recently published in the Journal of Proteome Research. Several lines of genetically-engineered mice have also been generated with the goal of developing new models for gastric cancer. Several strategies are being pursued to develop a model which develops tumors with high specificity in the stomach.